S1PR2 variants associated with auditory function in humans and endocochlear potential decline in mouse

Progressive hearing loss is very common in the population but we still know little about the underlying pathology. A new spontaneous mouse mutation (stonedeaf, stdf ) leading to recessive, early-onset progressive hearing loss was detected and exome sequencing revealed a Thr289Arg substitution in Sphingosine-1-Phosphate Receptor-2 (S1pr2). Mutants aged 2 weeks had normal hearing sensitivity, but at 4 weeks most showed variable degrees of hearing impairment, which became severe or profound in all mutants by 14 weeks. Endocochlear potential (EP) was normal at 2 weeks old but was reduced by 4 and 8 weeks old in mutants, and the stria vascularis, which generates the EP, showed degenerative changes. Three independent mouse knockout alleles of S1pr2 have been described previously, but this is the first time that a reduced EP has been reported. Genomic markers close to the human S1PR2 gene were significantly associated with auditory thresholds in the 1958 British Birth Cohort (n = 6099), suggesting involvement of S1P signalling in human hearing loss. The finding of early onset loss of EP gives new mechanistic insight into the disease process and suggests that therapies for humans with hearing loss due to S1P signalling defects need to target strial function.

Progressive hearing loss is the most common sensory deficit in the human population and can begin at any age, but we know very little of the underlying molecular or cellular mechanisms involved. Any genes found to be involved in progressive hearing loss can give insight into the molecular pathways and pathological processes leading to deafness. Mouse mutants are valuable tools to provide candidate genes as well as opportunities for detailed mechanistic investigation of the disease process.
Sphingosine-1-phosphate (S1P) signalling is known to be required for normal auditory function from studies of deaf mouse mutants. Mutation of the S1P transporter Spns2 leads to rapidly-progressive loss of auditory sensitivity 1 and three independent null mutations of the S1P receptor S1pr2 have severe elevations in auditory thresholds before 4 weeks old [2][3][4] . S1P is a lysophospholipid intermediate in the process of degradation of sphingolipids, but it also acts as a signalling molecule with effects both within the cell and outside 5,6 . Extracellular S1P signals through five different receptors, S1pr1-5, which in turn activate a variety of intracellular signalling pathways via G-proteins 7 . S1P signalling has been implicated in a range of functions including lymphocyte trafficking 6,8,9 , macrophage and mast cell function 10,11 , angiogenesis 12 , vascular permeability and tone 13 and bone remodelling 14 . It is not clear which if any of these functions is involved in normal hearing.
Here we report a missense mutation in S1pr2 that arose spontaneously in mice generated by a large scale targeted mutagenesis program 15,16 . We describe the identification of the causative mutation in the S1pr2 gene by linkage analysis and exome sequencing, the rapidly progressive hearing loss associated with a decline in the endocochlear potential, and subsequent loss of cochlear hair cells. Furthermore, we found that genomic variants close to the human S1PR2 gene were significantly associated with auditory thresholds in a large population

Results
As part of the Sanger Institute Mouse Genetics Project, new lines of mutant mice on a C57BL/6N genetic background are screened at 14 weeks old using an electrophysiological test of hearing, the Auditory Brainstem Response (ABR) 16,17 . In one line carrying a targeted disruption of the Mms22l gene (MMS22-like, DNA repair protein), most mice had normal ABR thresholds (n = 7), but others (n = 4) showed no response to sound stimuli, including one wildtype littermate used as a control. The deafness phenotype was isolated in a new mouse colony and showed transmission consistent with monogenic autosomal recessive inheritance with full penetrance. We named the mutant allele stonedeaf (stdf ).
The stdf mutation was mapped to a 10 Mbp interval of proximal chromosome 9 by linkage analysis of offspring from a [(stdf/stdf x C3HeB/FeJ)F1 x stdf/stdf ] backcross. DNA from two distantly-related affected mice was submitted for exome sequencing and the data filtered and analysed (Tables [1][2][3][4][5]. A large number of small structural variations were detected by Pindel across the whole genome. After filtering, 4 remained in the critical 10 Mbp region, but they were not located in coding regions, and were not confirmed by capillary sequencing. Two were present in the ancestral ES cell line (Table 3). Single nucleotide variants (SNVs) called by SAMtools were filtered as described (Table 4). Only 7 were within the critical 10 Mbp region (Tables 4 and 5  were in coding regions; one was synonymous (in Rdh8) and the other two resulted in amino acid changes. One of the two nonsynonymous SNVs (in Bmper) was found in the ancestral ES cell line used in the creation of the knockout from which the stonedeaf allele arose (Table 5) and the SNV (A > C substitution producing an amino acid change of T570P) was predicted to be neutral, so we considered this to be an unlikely candidate. The remaining SNV (a G > C transition in S1pr2; Fig. 1A), resulted in a threonine to arginine change (T289R), which was predicted by both NetDiseaseSNP 18 and PhD-SNP 19,20 to be deleterious. The affected threonine residue is highly conserved (Fig. 1B,D), located in a transmembrane domain, and was predicted by ConSurf 21,22 to be structural (Fig. 1C). Genotyping a large population (n = 204) of mice from the stonedeaf colony for this mutation by capillary sequencing supported its causative nature; all mice with raised thresholds (>4 weeks old) were homozygous for the mutation, whereas mice with normal thresholds were either S1pr2 +/stdf or S1pr2 +/+ . S1pr2 was subjected to protein structure prediction using Phyre2 23 , based on the X-Ray crystal structure of homologous receptor, S1pr1 24 and the T289R mutation was introduced using Pymol ™ (DeLano Scientific LLC) ( Fig. 1E-I). The mutation sits within a hydrophobic pocket and is likely to cause some local disruption including changing available hydrogen bonds (Fig. 1H,I). The mutation site (289T) is situated on the boundary between the membrane and intracellular region of the receptor. It might be anticipated that the replacement of threonine at position 289 with an arginine residue would have little effect on the protein's position within the membrane. However, membrane proteins are often highly sensitive to mutation and modest differences in the tilt angle of the transmembrane region relative to the membrane surface, or slight changes in insertion depth have been known to cause significant effects on function (several such cases have been reviewed 25 ). In S1pr2, the position of the snorkelling helix just above the aberrant arginine residue may be affected which might perturb intracellular signalling function (See Fig. 1E-G). The structural modelling therefore supports our conclusion that the T289R mutation is the stdf causal mutation. Immunohistochemistry was used at postnatal day (P)5 and P14 (Fig. 2) to reveal expression of S1pr2 mainly in the cell bodies of inner and outer hair cells (Fig. 2C,D), and in the stria vascularis and spiral ligament fibrocytes which are important for the recycling of ions into the scala media (Fig. 2B,F), and in spiral ganglion cells. The same pattern of expression was observed at P28 (data not shown). S1pr2 expression was detected with a similar distribution in the S1pr2 stdf/stdf mutants (Fig. 2E), which is not surprising as the point mutation is unlikely to stop protein translation.
Changes in auditory sensitivity with age were assessed by ABRs 17 , using capillary sequencing to assign genotype ( Fig. 3A-C). Mice homozygous for the stonedeaf mutation demonstrated a wide range of auditory sensitivity at 4 weeks old (Fig. 3A), while thresholds were higher at 8 and 14 weeks old ( Fig. 3A-C). For some of these mice   Table 5. Filtered SNVs identified in the critical region on chromosome 9. Details of the SNVs which passed the first five filtering steps in Table 4. They had a high mapping quality and read depth, were present in both mice sequenced, were not present in any wildtype strains or dbSNP128, and were within the mapped region. is conserved between mouse, human, xenopus and zebrafish. Note that S1pr2 is on the reverse strand, and so the G > C SNV appears as C > G when the sequence is oriented in the direction of reading of the gene, as in A and B in this figure. (C) ConSurf analysis of the residues; 289T is predicted to be buried (b) and highly conserved, and thus likely to be a structural residue (s). (D) ConSurf-annotated clustal of S1pr2 protein sequences. 289T is highly conserved. In (C,D), yellow lettering indicates insufficient data for a reliable calculation of conservation. (E) Structural model of the wildtype S1pr2 protein based on Phyre2 modelling from the S1pr1 structure. The seven transmembrane helices are shown in different colours. A black arrow points to the snorkelling helix (coloured red) referred to in the text. (F) A zoomed view of the wildtype 289T residue (green stick form) (G). A zoomed view of the Arginine residue (red stick form) in the T289R substitution of S1pr2 stdf . The effect of the mutation on the S1pr2 hydrogen bond network (based on Phyre2 modelling) is shown in H and I. (H) In the wildtype, T289 (purple) forms a single hydrogen bond (dotted line) with V286 (brown). (I) In the S1pr2 stdf mutation, R289 (purple) can no longer form a hydrogen bond with V286 (brown) or any other residues.
(n = 62), ABR measurements were made longitudinally at 4, 8 and 14 weeks old (Fig. 3D). A progressive deterioration in threshold with increasing age was apparent in S1pr2 stdf/stdf mice (n = 28). Wildtype (n = 7) and heterozygous (n = 27, not shown) mice maintained good sensitivity of threshold as age increased. Expression of S1pr2 in the lateral wall of the cochlear duct suggested a potential role for S1P signalling in generation of the high resting potential, the endocochlear potential (EP), in the endolymph bathing the upper surface of the sensory hair cells. EP is essential for normal hair cell function, providing a strong driving force across transduction channels. We measured EP in mice aged P14, P28 and P56, in some cases following ABR recording. In P14 mice, ABR thresholds of S1pr2 stdf/stdf mice were comparable to those recorded in S1pr2 +/stdf littermate controls (Fig. 4A). EPs recorded were also of comparable magnitude in S1pr2 stdf/stdf mice (98.0 ± 12.5 mV) and S1pr2 +/stdf littermate controls (100.0 ± 11.8 mV) (Fig. 4D). At P28, ABR thresholds of control mice were maturing towards adult levels, but homozygous mutants demonstrated a wide range of sensitivity (Fig. 4B). At P28, EP in control (heterozygote) mice had reached fully mature levels (125.5 ± 6.5 mV) but mutants demonstrated a reduced and variable EP (47.9 ± 22.2 mV) (Fig. 4E). At P56, ABR thresholds of mutants showed further elevations compared to littermate controls (Fig. 4C) and EP remained low (53.7 ± 17.9 mV). Littermate control heterozygotes maintained normal EPs at P56 (120.5 ± 9.8 mV) (Fig. 4F). In summary, at early stages of the pathological process, ABR thresholds increase in line with the reduction in EP.
The stria vascularis on the lateral wall of the cochlear duct is the major site of EP generation, through active pumping of K + ions into the endolymph. As EP was reduced in mutants, we examined the stria vascularis in whole mount preparations labelled with isolectin GS-IB4 to highlight capillaries and phalloidin to indicate filamentous actin at the boundaries of the marginal cells lining the luminal surface of the stria 1 . We examined strial samples from mice aged P14, P28 and P56. Strial marginal cell boundaries appeared normal in stonedeaf homozygotes at P14 (Fig. 4G,H). However, at P28, these boundaries were less regular in mutants than in heterozygous controls, showing a mixture of larger and smaller boundaries, typically losing their normal hexagonal/ pentagonal shapes (Fig. 4I,J). These changes were seen intermittently with some normal segments in between. This structural defect progressed with age. At P56, some extremely large or small marginal cell boundaries were seen (Fig. 4K,L). Strial capillaries displayed clear dilation in patches in some of the mutants, which was not seen in any of the controls (Fig. 5A-F).
Kcnj10 is an ATP-sensitive inward rectifying K + channel which plays a crucial role in the movement of K + into the endolymph to establish the high [K + ] and positive EP 26 , so we looked for expression in the stria vascularis of stonedeaf mutants at P5 and P28. Expression appeared as strong in mutants as in controls at P5 but by P28, immunolabelling was weaker in the basal turn of mutants compared with controls ( Fig. 5G-J).
No gross anatomical defects were found in the middle ear, ossicles or inner ear in stonedeaf mutants. The organ of Corti was then examined by scanning electron microscopy. All mice were ABR-tested one week before fixation and later genotyped by sequencing. In five-week-old mutants, some abnormalities of the organ of Corti (D,E) S1pr2 cochlear expression at P14 in stonedeaf mutant (S1pr2 stdf/stdf ) and littermate controls (S1pr2 +/stdf ). We detect expression of S1pr2 mainly in spiral ligament fibrocytes (red arrowheads), OHCs (bracket), individual cells in the stria vascularis (black arrowhead), basilar membrane (red arrows) spiral ganglion cells (arrows) and spiral ganglion neuron terminals below the IHCs (green arrowheads). We did not detect a difference in S1pr2 expression distribution, or any obvious difference in labelling intensity, between stonedeaf mutants and littermate controls. F: Expanded view of S1pr2 expression in the spiral ligament fibrocytes (red arrowheads). Scale Bars: (A,D-F) 20 μm. (B,C) 10 μm.
were noted only in severely impaired S1pr2 stdf/stdf mice. Homozygotes with click thresholds of 95dB SPL, or with no ABRs at all, showed OHC degeneration from the middle turn of the cochlea towards the base, while homozygotes with click thresholds at or better than 80dB SPL appeared to have largely normal organs of Corti (Fig. 6). Degeneration of hair cells was much more extensive in 9-week-old affected mice. Only the extreme apical 10-20% of the organ of Corti retained some inner and outer hair cells in S1pr2 stdf/stdf mice at this age and there was extensive degeneration in more basal regions (data not shown). A base-to-apex gradient in hair cell degeneration is a common pathological pattern.
We next asked if the targeted Mms22l mutation had any impact upon the auditory phenotype. Following identification of the stdf mutation, mice with ABR data were divided according to both the S1pr2 genotype and the Mms22l genotype (n = 73 mice aged 4 weeks, 118 mice aged 8-10 weeks, 14 mice aged 44 weeks). However, the presence or absence of the targeted Mms22l allele did not significantly affect ABR thresholds within any S1pr2 genotype and age group (Mann-Whitney Rank Sum Test, p > 0.05 in all cases).
To assess the possibility that other phenotypes were associated with the S1pr2 stdf mutation, we obtained DNA samples from 39 mice from the Mms22l line that had been screened as part of the Mouse Genetics Project 16 . These samples were genotyped for the S1pr2 stdf mutation and mice grouped into new cohorts that were wildtype (n = 10), heterozygous (n = 17) and homozygous (n = 12) for the S1pr2 stdf mutation. Phenotyping data for 252 parameters measured across 22 broad assay groups 16 were obtained for these 39 mice and analysed to determine significant phenotypes. A summary of these data are shown in Fig. 7 as a heatmap, with data for the stonedeaf mutants alongside that of the Mms22l mutants and of Spns2, another mutant line affecting an S1P transporter protein involved in the same signalling pathway and with a similar auditory phenotype 1 . The S1pr2 stdf mutation had a selective effect only on hearing, and showed no other abnormalities over the range of phenotyping tests (the red box in Fig. 7 indicating abnormal neurological features results from a lack of Preyer reflex, the ear flick response to sound, which is part of the neurological assessment). The targeted Mms22l allele led to lethality in homozygotes, so heterozygotes were screened but no other anomalies were detected other than the three individuals that were deaf and later proved to be homozygous for the stonedeaf mutation. In contrast, the Spns2 mutants showed deafness and a number of additional abnormal phenotypes (Fig. 7), including eye defects 1 and various anomalies of the immune system, especially associated with leucocytes 9 . S1P signalling effects on the immune system and on retinal angiogenesis are mediated through the S1pr1 gene (for example [27][28][29][30] ) and similar deficiencies would not be expected in the S1pr2 stdf mutant mouse.
Finally, we investigated the potential involvement of S1PR2 in human hearing loss by asking if variants in this gene show association with auditory thresholds in the 1958 British Birth Cohort 31 . Hearing thresholds measured at 1 and 4 kHz at the age of 44-45 and genotype data imputed to the 1000 Genomes dataset (March 2012 release) were used in this association study. Both datasets were available for 6,099 individuals. SNPs within 0.1 Mb of the S1PR2 gene as well as within the gene itself were interrogated as a candidate gene association. For 1 kHz audiometric thresholds, the adjacent SNP rs74930654 (which is 27 kb from the end of S1PR2, on the reverse strand) showed the most significant association, with a p value of 0.001644. For 4 kHz thresholds, the most significant association was with rs201930568 (located 100 kb before the start of S1PR2, on the reverse strand), with a p value 0.001105. These findings suggest that variants affecting the S1PR2 gene do contribute to auditory thresholds in the UK population.
stonedeaf mice appeared to be a secondary phenomenon, occurring after the onset of pathological changes to the stria vascularis, EP and ABR thresholds. Indeed, normal-looking hair cells were seen in the apex of the cochlea in 5-and 9-week-old mice with no measureable ABRs. Secondary hair cell degeneration resulting from a reduced EP has been described before in mouse mutants where (unlike S1pr2) there is no expression of the mutant gene in the hair cells, such as in mice with mutations of the Kit gene 32 . A normal EP level may be critical for the maintenance of the sensory hair cells, which are specialised to survive in an environment where their upper surface is bathed in endolymph, so any change in voltage or ionic composition of endolymph may disrupt their homeostasis. S1pr2 is expressed in hair cells so we cannot rule out a direct effect of the mutation on hair cells, but as the earliest cochlear defect we saw was the reduction in EP, this is likely to be the primary pathological event leading to raised ABR thresholds and secondary hair cell degeneration.
(C) At P56, ABR thresholds of S1pr2 stdf/stdf mutants (n = 14) showed further elevations compared to littermate controls (n = 7). (F) Mean EP from P56 mutant mice remained low (n = 20; 53.7 ± 17.9 mV, range 1.6-84.2 mV) while S1pr2 +/stdf control mice maintained normal EPs (n = 12; 120.5 ± 9.8 mV, range 91.8-130.9 mV). (K,L) Strial marginal cell boundaries became more irregular at P56 in mutants. Figure 5. Capillary expansion and Kcnj10 expression in the stria vascularis of S1pr2 stdf mice. Capillaries in the middle turn stria vascularis of mutants, shown for S1pr2 stdf/stdf mutants (B,D,F) and in S1pr2 +/+ littermate controls (A,C,E) at P14 (A,B), P28 (C,D) and P56 (E,F). Capillaries appeared swollen at P28 and P56 in S1pr2 stdf/stdf mutants (D) compared to S1pr2 +/+ littermate controls (C). Panels A-F were imaged at the same magnification; scale bar (in F): 20 μm. Kcnj10 (brown label) was expressed at P28 in the stria vascularis intermediate cells in both S1pr2 stdf/stdf mutants (H,J) and in S1pr2 +/+ littermate controls (G,I), in the apical and basal turns of the cochlea. Labelling appeared strongest towards the apex (G,H) and less strong towards the base (I,J). Scale Bars (G-J): 10 μm. Sections of the entire inner ear were examined, n = 3 mice for each genotype. Alternatively, the dilation we see may be a consequence of abnormal strial function, leading to both reduced EP and compensatory changes of the vasculature in response to local metabolic demand. Previous reports of mutants with strial dysfunction due to a lack of intermediate cells suggest that the strial vasculature responds to metabolic demand leading to small capillaries where the stria is not functional 34 .
The pattern of reduced EP and altered strial morphology early in the progression of the hearing loss is very similar to that seen in mice with a mutation of Spns2, another component of S1P signalling 1 . A progression to fewer, larger marginal cells has been described previously in ageing BALBc and CBA/CaJ mice as their EP becomes lower 35,36 and in other mutant mice with EP defects such as the pendrin mutant 37 . Three independent knockout alleles of S1pr2 have been reported previously [2][3][4] , and disrupted marginal cell boundaries and expanded strial capillaries were described in one of these mutant lines 3 although no EP measurements were made. These knockout mice all showed more severe hearing impairment than we see in the stonedeaf mutant, although none of them reported thresholds as early as P14 so it is not clear if there was normal development followed by very rapid loss of sensitivity. We find thresholds are normal in S1pr2 stdf homozygotes at P14, and by P28 there is considerable variability in thresholds despite all the mice having the same genetic background (C57BL/6N) and being housed in identical environmental conditions. In two targeted null alleles of S1pr2, severe hearing impairment was reported in all mice tested at P22 2 or 4 weeks old 3 , so the stonedeaf missense mutation may produce a hypomorphic effect on S1pr2 function and slower progression of the phenotype. The three null alleles of S1pr2 were also reported to show severely abnormal stria vascularis morphology, vestibular defects and progressive spiral ganglion loss. We did not see any indications of vestibular defects in S1pr2 stdf homozygotes, but we did find stria vascularis structural defects that progressed with age and were similar to those previously reported in a null allele 3 .
In conclusion, we report here for the first time that EP is reduced resulting from an S1pr2 mutation, a feature that was not reported in the three knockout S1pr2 mutants. Furthermore, we found that the S1pr2 stdf homozygotes show normal thresholds and EP around the onset of hearing, progressing over the following few weeks to severe or profound deafness, greatly reduced EP and presumed secondary hair cell degeneration. The delayed loss of auditory function seen in S1pr2 stdf homozygotes suggests that the functional processes mediated through S1P-signalling are not critical for the development of hearing, but instead are required for the ongoing maintenance of cochlear function once hearing has been established. This suggests the possibility of therapeutic intervention to prevent the hearing loss if diagnosed early in the pathological process. This is particularly relevant Figure 6. ABR data from individual mice with corresponding scanning electron micrographs. ABR thresholds from one S1pr2 +/stdf mouse (A) and two S1pr2 stdf/stdf mice (E,I) recorded at four weeks with scanning electron micrographs from the same mice at five weeks old. (B-D) show the apical, middle and basal turns respectively of the S1pr2 +/stdf mouse; the corresponding best frequency for each region illustrated is indicated on the audiogram (A). (F-H) likewise show the apical, middle and basal turns of the S1pr2 stdf/stdf mouse whose audiogram is shown in (E), with the corresponding frequencies marked. Although this mouse has raised thresholds, the organ of Corti appears normal. (J-L) show the apical, middle and basal turns of a S1pr2 stdf/stdf mouse with no response to stimuli at all (I). No hair cells are visible in the basal turn (L), but some remain in the mid-apical (K) and apical (J) turns. Scale bar = 10 μm.
given the recent discovery of missense mutations in human S1PR2 38 and our discovery of significant association between genomic markers close to the S1PR2 gene and auditory thresholds in the human population. These findings emphasize the need for accurate diagnosis and development of treatments to restore EP in such cases where S1P signalling is implicated; simply attempting to regenerate hair cells, for example, would not restore hearing if a patient has a pathology involving EP decline. Discussion and research on developing treatments for hearing loss often focus on sensory hair cells, but our report provides the first clear example of the involvement of S1PR2 in EP generation and the need to restore EP if treatments are to be successful.

Materials and Methods
Ethics statement. Mouse studies were carried out in accordance with UK Home Office regulations and the UK Animals (Scientific Procedures) Act of 1986 (ASPA) under UK Home Office licences, and the study was approved by the King's College London Ethical Review Committees. Mice were culled using methods approved under these licences to minimize any possibility of suffering.
Origin and identification of mutant allele. The stonedeaf (stdf ) mutation arose in mice derived from the Sanger Institute Mouse Genetics Project, which uses targeted ES cells to generate new mutant lines as described previously 15,16 . The mutation was mapped by linkage analysis of offspring from a [(stdf/stdf x C3HeB/FeJ)F1 x stdf /stdf ] backcross using ABR thresholds at 8-10 weeks old to assign phenotype (n = 79 affected and 82 unaffected offspring). DNA from two distantly-related affected mice was submitted for exome sequencing using the Agilent SureSelect XT mouse all exon kit for sequence capture, sequenced on the Illumina HiSeq 2000 platform and the data filtered and analysed as specified in Tables 1-5. The sequences have been deposited at ENA, accession number ERP000739. Mutant mice are available through EMMA.
Modelling the mutation. S1pr2 was subjected to protein structure prediction using Phyre2 (Protein Homology/analogY Recognition Engine version 2.0 23 ). The coordinates corresponding to the top hit, which was based on the X-Ray crystal structure of homologous receptor, S1pr1 24 , were used to produce a model of S1pr2 and to introduce the T289R mutation using Pymol ™ (DeLano Scientific LLC).
Electrophysiology. Auditory sensitivity was assessed using measurements of Auditory Brainstem Responses (ABRs) 17 . ABRs were recorded from Ketamine/Xylazine anaesthetised mice aged 4, 8 and 14 weeks, evoked by clicks and tone pip (6-30 kHz) stimuli presented freefield at 0-95 dB SPL in 5 dB steps to determine threshold sensitivity. EPs were measured under urethane anaesthesia (2 mg/g) in a different group of mice, in some cases following ABR threshold measurement. After these ABR recordings, a tracheal cannula was inserted and the bulla was opened to allow fenestration of the cochlear wall and insertion of a 150 mM KCl-filled glass micropipette into the scala media. The EP was measured as the potential difference between the scala media and a reference silver/ silver chloride pellet under the dorsal skin of the neck 39 .
Scanning electron microscopy. Samples were taken from S1pr2 +/stdf and S1pr2 stdf/stdf mice aged 5 and 9 weeks old, a week after ABR testing at 4 and 8 weeks old respectively, and processed for scanning electron microscopy of the apical surface of the organ of Corti using the OTOTO methodology 40 . Samples were analysed according to their percentage distance from the cochlear base.